Inwardly rectifying potassium channels influence Drosophila wing morphogenesis by regulating Dpp release

Dominant-negative mutations in the gene encoding the Kir2.1 potassium channel (also known as Irk) lead to Andersen-Tawil syndrome (ATS), a condition comprising diverse symptoms including craniofacial and digit patterning defects, periodic paralysis, cardiac arrhythmias and cognitive deficits (Plaster et al., 2001; Tristani-Firouzi et al., 2002; Yoon et al., 2006). Periodic paralysis, cardiac arrhythmias and cognitive deficits are likely consequences of abnormal Kir2.1 function in neurons and muscle. By contrast, it is not obvious how Kir2.1 channels regulate craniofacial and limb development.

In non-human model systems, ion channels also play important developmental roles (Bates, 2015; Levin, 2014a,b). For example, disruption of Kir2.1 function in mouse, frogs and flies results in defects in morphogenesis, suggesting a conserved developmental role for inwardly rectifying K⁺ channels (Adams et al., 2016; Dahal et al., 2012; Zaritsky et al., 2000). In mice, vascular development and remodeling requires a mechanically gated cation channel (Ranade et al., 2014). In chick limb chondrocytes, inhibition of potassium channels alters the pattern of spontaneous calcium transients and reduces the percentage of cells that differentiate into cartilage (Varga et al., 2011). Further, ion channel function is necessary for dorsal closure during early embryogenesis in Drosophila (Hunter et al., 2014). Thus, ion channels play an essential role in morphogenesis. However, the underlying mechanisms remain unclear.

Previously, we demonstrated that deletion of Kir2.1 in mouse resulted in ATS-like phenotypes that were reminiscent of those produced by disruption of the bone morphogenetic protein (BMP) signaling pathway (Dahal et al., 2012). BMPs are members of the transforming growth factor beta (TGFβ) superfamily. We complemented the mouse studies with experiments in Drosophila because flies express Kir2.1 orthologs (Irk1, Irk2 and Irk3), and the role of a BMP ortholog, Decapentaplegic (Dpp), in the developing wing is well characterized. We showed that reducing the function of any Irk subunit in the wing produces phenotypes that resemble those produced by reduced Dpp signaling, such as held-out wings, thickened veins, venation pattern defects, small wings, and ablation of almost the entire wing (Adachi-Yamada et al., 1999; Chen et al., 1998; Dahal et al., 2012; de Celis, 1997; Letsou et al., 1995; Ralston and Blair, 2005; Shimmi et al., 2005; Spencer et al., 1982; Zecca et al., 1995; Zeng et al., 2007). Irk channels can be heteromeric, made up of subunits encoded by different genes (Bichet et al., 2003; Hibino et al., 2010). All three Irk subunits are expressed in the wing disc and loss of one channel subunit leads to compensatory increased expression of the other two (Dahal et al., 2012). Therefore, to mimic ATS and to eliminate the function of native Irk channels, but not other potassium channels, we expressed a dominant-negative Irk subunit, Irk2DN, in the wing. Irk2DN expression led to severe wing phenotypes, similar to those produced by dpp^d8/dpp^d10 flies that selectively reduce Dpp function in the developing wing (Zecca et al., 1995). Moreover, deletion, mutation or RNAi knockdown of Irk2 function led to reduced phosphorylation of Mad (the Dpp pathway-specific intracellular mediator Smad protein) and reduced Dpp target gene expression (Dahal et al., 2012). Although this study revealed that Dpp signaling requires Irk channel function, it did not identify the underlying mechanism by which a potassium channel could influence Dpp signaling. Here, we provide evidence that loss of Irk function results in abnormal release of Dpp from the producing cell, the first step in the intracellular signaling pathway.

The Dpp signaling pathway works with other signaling pathways to establish the pattern of the fly wing (Burke and Basler, 1996; Capdevila and Guerrero, 1994; Posakony et al., 1990). In the primordial wing disc, cells along the anterior/posterior (A/P) compartment boundary produce Dpp (Nellen et al., 1996). Release of Dpp elicits a patterning response in neighboring cells by binding of Dpp to the type-2 kinase receptor Punt, which then phosphorylates a type-1 kinase receptor, Thickveins (Tkv) (Letsou et al., 1995; Nellen et al., 1994; Raftery and Sutherland, 1999). Phosphorylated Tkv (p-Tkv) phosphorylates Mad (Wiersdorff et al., 1996). Phosphorylated Mad (p-Mad) then forms a complex with a co-Smad (Medea), enters the nucleus, affects the transcription of Dpp target genes and thereby regulates cell fate (Maduzia and Padgett, 1997; Newfeld et al., 1997; Raftery and Sutherland, 1999; Raftery et al., 1995; Sekelsky et al., 1995; Wiersdorff et al., 1996).

Multiple models have been proposed to explain how the extracellular distribution of Dpp along the A/P boundary affects signaling and consequent patterning. One model proposes that restricted diffusion of Dpp forms an extracellular gradient that designates cell fate according to Dpp concentration. GFP-tagged Dpp (Dpp-GFP) forms a gradient that mimics anti-p-Mad staining from the A/P boundary across the wing disc (Raftery and Umulis, 2012). Increasing the expression of Dpp increases the expression of Dpp target genes (Nellen et al., 1996), supporting the hypothesis that cellular fate within the wing depends on the concentration of Dpp that a cell receives (Nellen et al., 1996; O'Connor et al., 2006). More recently, the lab of Thomas Kornberg has shown that Dpp travels along specialized long filopodia (cytonemes) that synapse upon responding cells (Roy et al., 2014). Genetic disruption of cytonemes perturbs Dpp signaling (Roy et al., 2014). Thus, these data support a second model in which Dpp elicits its patterning response through contact-mediated signaling (Roy et al., 2014). Another recent concept is that the temporal pattern of presentation of Dpp-like ligands is also important for downstream signaling and responses (Sorre et al., 2014). In cell culture, pulsatile presentation of TGFβ doubles the TGFβ target transcriptional response compared with that produced by constant exposure to the same concentration of ligand (Sorre et al., 2014). The temporal dynamics of ligand presentation is important in other types of contact-mediated signaling, such as the signaling between neurons.

Here, we use the fly wing disc model to obtain mechanistic insight into Irk-dependent regulation of Dpp signaling in vivo. We find that Irk channels are required in the Dpp-producing cells. By monitoring release of Dpp with a fluorescently tagged version of Dpp, we record Dpp release events in vivo. Upon inhibition of Irk channels, the baseline fluorescence is increased and the incidence of discrete Dpp-GFP release events above baseline is reduced. Overall, our results support a model in which Irk function plays an essential role in the regulated release of Dpp and resultant tissue morphogenesis.

Irk channels regulate vesicle release in neurons and pancreatic beta cells by modulating membrane potential to influence intracellular calcium dynamics. We hypothesized that Irk channels modulate Dpp release by a similar mechanism. Calcium imaging reveals transient elevation of intracellular calcium in individual wing disc cells. Inhibition of Irk channels decreases the duration and amplitude of the calcium transients and alters Dpp release dynamics, suggesting a possible connection between intracellular calcium and Dpp release. We show that high extracellular K⁺ concentrations induce Dpp release from cultured wing discs, suggesting that depolarization promotes release of Dpp. Together, our results indicate that Irk channels regulate release of Dpp and that the temporal dynamics of Dpp presentation impacts the downstream transcriptional response and resultant tissue morphogenesis.

Our previous work demonstrated a role for Irk channel function in Dpp-mediated wing morphogenesis (Dahal et al., 2012). However, this previous study did not provide a cellular or mechanistic understanding of how Irk channels regulate wing disc patterning. Irk2 belongs to the Kir2 subfamily of inwardly rectifying K⁺ (Kir/Irk) channels that act to maintain a stable resting membrane potential (Döring et al., 2002; Hibino et al., 2010; Luan and Li, 2012; MacLean et al., 2002). To test whether the effects produced by Irk2DN reflect a requirement for the inwardly rectifying property of the channel, we assayed for rescue of Irk2DN-induced defects by transgenic expression of different classes of Kir genes. Kir transgenes were tagged with green fluorescent protein (GFP) or Flag to allow independent confirmation of expression and membrane localization (Fig. S1). We used a wing-specific driver, MS1096-Gal4 (MS1096>), that alone has no effect on wing size but can lead to mild defects (bristle transformations or mild wing venation defects) (Fig. 1A,I).

Expression of Irk2DN, which acts similarly to human ATS alleles that block Kir2 channel activity, results in severely defective rudimentary wings that consist of only the hinge region, wings of reduced size or with thickened vein tissue (Fig. 1B,C,I) (see also Dahal et al., 2012). These rudimentary wings reflect apoptosis in the late third instar larval wing disc, similar to the phenotype that results from loss of Dpp signaling (Dahal et al., 2012; de Celis, 1997, 2003). When apoptosis is blocked by expression of p53, we find that p-Mad is still reduced in those discs, indicating that reduction of Dpp signaling is upstream or independent of apoptosis (Dahal et al., 2012). Co-expression of human KIR2.1 (KCNJ2-GFP) with Irk2DN results in wings similar to those found upon expression of the driver alone (Fig. 1D,I). Thus, human KIR2.1 completely rescues the wing phenotypes produced by Irk2DN. Co-expression of another Kir channel type, rat Kir4.1 (Kcnj10-flag) (Hibino et al., 2010), with Irk2DN partially rescues the defects caused by inhibition of native Irk channels. Eleven percent of MS1096>Irk2DN; Kcnj10-flag wings had no detectable defects (Fig. 1E,I), whereas 35% had venation defects and 53% were severely defective (Fig. 1F,I). However, co-expression of another potassium channel type [Ork (Ork1 – FlyBase), a two-pore potassium channel], did not rescue the Irk2DN wing phenotype (Fig. 1H,I). Expression of Ork alone did not cause severe defects without expression of Irk2DN (Fig. 1G). Overall, two different mammalian Kir subfamily members, KIR2.1 and Kir4.1, rescue wing defects produced by expression of Irk2DN. KIR2.1 and Irk2 are both classified in the same subfamily of inwardly rectifying K⁺ channels (Döring et al., 2002; Hibino et al., 2010) and KIR2.1 rescues MS1096>Irk2DN wing phenotypes more effectively than a member of a different subfamily, Kir4.1. These results support the view that the inwardly rectifying property of these K⁺ channels plays a role in regulation of wing disc morphogenesis.

Dpp is released from the producing cell and then binds to tyrosine kinase receptors on the responding cell to trigger a signaling cascade that ultimately induces target gene expression. An initial step in the signaling cascade is phosphorylation of Mad, and p-Mad can be used to as a readout of Dpp signaling. Irk2DN expression diminishes p-Mad, indicating a defect in Dpp signaling (Fig. 2C,E,G) (see also Dahal et al., 2012). Loss of p-Mad could be due to loss of Dpp (the signal from producing cells) or its transduction (the signaling cascade in responding cells).

To distinguish between these possibilities, we used a genetic approach and expressed a constitutively active Dpp receptor, Tkv (tkvQD), with or without Irk2DN and assayed for p-Mad. Again, we used the MS1096 driver that results in broad expression in the center of the wing pouch (Fig. 2A). Transgenic expression of tkvQD in the wing pouch results in a strong p-Mad signal throughout the wing disc (Fig. 2D,G, n=20). However, co-expression of Irk2DN had no effect on the intensity of the p-Mad signal (Fig. 2F,G, n=22), suggesting that Irk channels act upstream of Mad phosphorylation in the Dpp pathway. As further evidence that Irk2 is not required in Dpp-responding cells we created clones that express Irk2DN in the Dpp-responding cells and these showed no change in p-Mad fluorescence intensity compared with neighboring cells (Fig. 2H-K, Fig. S2, n=12 discs and 38 clones). These data support the view that Dpp-responding cells do not require Irk channels in a cell-autonomous manner for phosphorylation of Mad, a crucial effector of Dpp signaling.

We next tested if Irk channels function in the Dpp-producing cells for Dpp signaling. We expressed Irk2DN in the cells that produce Dpp at the A/P boundary of the wing disc (Fig. 2B) using the dpp-gal4 (dpp>) driver and assessed p-Mad levels in dpp>Irk2DN and control (dpp> alone) wing discs. Inhibition of Irk channels in the Dpp-producing cells significantly reduced p-Mad in the center of the pouch of the wing disc (Fig. 2L-N, n=15 discs of each genotype averaged in Fig. 2L, P<0.005) demonstrating that Dpp-producing cells require Irk channel function for efficient presentation of the Dpp signal.

We next asked whether Irk channels function at the level of Dpp transcription, translation, processing or release. We first assessed the effects of Irk2DN on dpp transcription using a dpp-lacZ reporter. Irk2DN expression had no effect on dpp-lacZ expression (n=10 for both genotypes), nor did it affect expression of the Dpp receptor Tkv (Fig. S3, n=9 MS1096>tkv-lacZ and n=10 MS1096>Irk2DN; tkv-lacZ wing discs). We assessed effects of Irk2DN expression on intracellular Dpp protein levels using an antibody that recognizes the Dpp prodomain (Akiyama and Gibson, 2015), the portion of the Dpp proprotein that is cleaved prior to secretion. Dpp prodomain immunoreactivity in dpp>Irk2DN wing discs is not significantly different from that of dpp> wing discs (Fig. 3A-C, n=19 and n=18, respectively). This indicates that Irk2DN does not affect Dpp translation or levels of Dpp protein expression, at least as detectable by the prodomain antibody.

Next, we assayed for extracellular Dpp. Previous work (Entchev et al., 2000; Teleman and Cohen, 2000) demonstrated that the fluorescence intensity of Dpp-GFP in Dpp-producing cells provides an assessment of Dpp protein expression and distribution. Accordingly, we expressed Irk2DN along with dpp-GFP (dpp>Irk2DN; dpp-GFP) in the Dpp-producing cells (dpp>) and compared the GFP immunoreactivity with that of dpp>dpp-GFP wing discs. For this experiment, we prepared wing discs under conditions that do not permeabilize cellular membranes, so that the antibody would have access to extracellular but not intracellular Dpp-GFP (see Materials and Methods). Under these conditions, we find that inhibition of Irk channels in the Dpp-producing cells increases the peak fluorescence intensity and widens the A/P domain of extracellular Dpp-GFP (Fig. 3D-F, P=0.001 and 0.007, respectively, two-tailed t-test; n=27 dpp>dpp-GFP and n=33 dpp>Irk2DN; dpp-GFP wing discs). Taken together, the results suggest that Irk2DN expression increases the amount of extracellular Dpp and the area over which it localizes. This raises a paradox as to how to reconcile this conclusion with our data showing decreased Dpp signaling as seen by decreased p-Mad, decreased Spalt expression (Dahal et al., 2012) and by the wing patterning phenotypes (Figs 1 and 2). However, we did note that Irk2DN expression significantly increased the size of wing discs, which is a Dpp gain-of-function phenotype (Fig. 3G-J, Fig. S4). This could suggest a difference in the mechanism of Dpp signaling for patterning compared with wing disc size.

A possible technical explanation is that inhibition of Irk channels compromises membrane integrity, making cells more permeable to antibodies. To test this, we performed experiments under conditions that allowed simultaneous analysis of both intracellular and extracellular Dpp-GFP. Consistent with our measures of extracellular Dpp-GFP alone, the combined levels of intracellular and extracellular Dpp-GFP fluorescence also increase in intensity and area (Fig. 3G-K). This suggests that the increased distribution and intensity of Dpp-GFP fluorescence in Irk2DN-expressing wing discs are not due to a difference in membrane permeability. We then considered the possibility that Irk channels regulate the release of Dpp.

To test the possible function of Irk channels in the regulation of Dpp release, we used live imaging methods in combination with expression of Dpp-GFP. These methods take advantage of the pH dependence of GFP fluorescence intensity (Bokman and Ward, 1981; Kneen et al., 1998; Miesenbock et al., 1998; Ward and Bokman, 1982). The acidic environment of the vesicular lumen quenches GFP fluorescence (Gandhi and Stevens, 2003; Paroutis et al., 2004). By contrast, the neutral environment of the extracellular space, encountered by GFP upon exocytosis, restores GFP fluorescence (Fig. 4A, schematic). Thus, we monitored changes in fluorescence intensity over time to track the vesicular release of Dpp-GFP from Dpp-producing cells. First, we performed an important control to observe the unquenching of intracellular GFP: neutralizing the acidic environment of the lumen by treatment with ammonium chloride caused the Dpp-GFP fluorescence intensity to rapidly and significantly increase (Fig. S5A,B).

Next, we obtained time-lapse images of Dpp-GFP to monitor the spatial and temporal dynamics of GFP fluorescence intensity. In each dpp>dpp-GFP wing disc imaged we observed spontaneously appearing bright puncta of Dpp-GFP and then the fluorescence decreased in the brightest pixel of the central puncta, with a concomitant increase in fluorescence intensity of the surrounding concentric pixels (Fig. 4, Fig. S5C). The average decrease in fluorescence intensity of the central bright puncta (putative site of exocytosis) roughly equaled the sum of the increases in fluorescence intensity occurring in surrounding concentric pixels (Fig. 4B, Fig. S5C). In other words, the Dpp-GFP fluorescence spread with time consistent with GFP diffusing away from a release site, as expected for vesicular release of Dpp-GFP. Puncta that reached a threshold fluorescence above baseline were identified as discrete individual release events (Fig. 4B, Fig. S5C).

To determine whether Irk channels regulate Dpp release, we compared the changes in Dpp-GFP fluorescence intensity over time in the presence or absence of native Irk channel function. The average frequency of release events at any one particular release site in control wing discs was 1.7±0.07 events per 2 min and within the analyzed region there were 26±5 release events. Irk2DN expression reduced the incidence of discrete release events, with an average of 1.4±0.06 events per 2 min at individual release sites and 14±3 release events per analyzed region (Fig. 4B-E). However, Irk2DN increased the baseline fluorescence, quantified as the number of pixels with midrange fluorescence intensity (Fig. 4F), for cells located at either of the two wing disc areas analyzed [the dorsal-ventral (D/V) boundary region or the ventral edge]. There was no change in the number of pixels above high fluorescence intensity thresholds (Fig. 4G).

Overall, these data suggest that Dpp-GFP is released in discrete bursts in control wing discs. Loss of Irk2 channel function increases the baseline level of Dpp-GFP while at the same time decreasing the number of discrete high-threshold release events. This could reflect an increase in the number of release events containing smaller amounts of Dpp-GFP.

In excitable cells, inhibition of Irk channels alters resting membrane potential and thereby can affect intracellular Ca²⁺ and vesicle release (Bondy et al., 1987; Braun et al., 2008; Cilz et al., 2014; Rorsman and Braun, 2013). To determine whether wing disc cells exhibit changes in intracellular calcium levels, we expressed GCaMP6-slow (GCaMP6s) with or without Irk2DN in the Dpp-producing cells (dpp>Irk2DN; GCaMP6s and dpp>GCaMP6s, respectively) and then imaged GCaMP6s fluorescence in wing discs over time. In the dpp>GCaMP6s control, we found that calcium waves propagate through the cells of the wing disc (Fig. 5A). In addition, individual dpp>GCaMP6s cells of the Dpp-producing zone undergo transient elevations of intracellular Ca²⁺ (Fig. 5C, top). However, in dpp>Irk2DN; GCaMP6s wing discs, calcium transients of both types are of reduced amplitude and are shorter in duration, in contrast to discs that express GCaMP6s alone (n=8 discs for each genotype, Fig. 5B-D). These data show that inhibition of Irk channels alters spontaneous calcium transients. Overall, Irk activity regulates the amplitude and the duration of calcium spikes in these epithelial cells of the developing embryo. To determine if depolarizing conditions can induce calcium transients, we applied high extracellular K⁺ to cultured wing discs while live imaging. We found an increase in GCaMP6s fluorescence that corresponded to the time at which the K⁺ solution was added, indicating that depolarization can induce calcium transients (Fig. 5E-G).

Because Irk channels regulate membrane potential in several cell types (Hibino et al., 2010; Vaca et al., 1996), we hypothesized that the membrane potential may regulate the dynamics of Dpp release. To test this hypothesis, we imaged Dpp release in wing discs while we added a K⁺ solution to depolarize the cells. For this assay, we used wing discs expressing Dpp tagged with a highly pH-sensitive GFP variant, Dpp-super ecliptic phlorin (Dpp-SEP), which is less fluorescent at low pH than GFP and therefore has a better signal to noise ratio. Addition of K⁺ solution led to an increase in the fluorescence of Dpp-SEP in dpp>dpp-SEP wing discs (Fig. 5H-J). The ability of high K⁺ to trigger elevations of intracellular calcium as well as the release of Dpp-SEP implicates membrane potential as a regulator of these physiological processes. Further, these data suggest that Irk channels regulate Dpp release through impacting membrane potential (Fig. 6).

Irk channels regulate resting membrane potential in excitable cells and thus can modulate how readily a neuron fires an action potential, a muscle contracts, or a pancreatic beta cell secretes insulin (Bondy et al., 1987; Braun et al., 2008; Cilz et al., 2014; Hibino et al., 2010; Riz et al., 2015; Rorsman and Braun, 2013; Vaca et al., 1996). A mechanism explaining how Irk channels could impact morphogenesis was less clear. The results reported here provide insights into the underlying mechanism. First, Irk2DN-induced wing defects are most efficiently rescued by KIR2.1, a strong inward rectifier, whereas Kir4.1, a moderate inward rectifier, partially rescues wing phenotypes (Fig. 1). By contrast, an open rectifier potassium channel does not rescue Irk2DN wing phenotypes. We conclude that the inward rectifying property is important for the developmental function of Irk channels. The inward rectifying property is important for setting resting membrane potential, and when this type of channel is inhibited in other cell types they become depolarized. Second, we show that Irk channels function in the Dpp-producing cells and are not required cell-autonomously for transmission of the Dpp signal (Fig. 2). Third, we show a difference in Dpp distribution and the dynamics of Dpp release in Irk2DN compared with control wing discs (Figs 3 and 4). Fourth, we show that inhibition of Irk channels decreases the duration and amplitude of native calcium activity (Fig. 5). Finally, we show that depolarization with extracellular potassium can induce Dpp release (Fig. 5). Together, these data lead us to hypothesize that Irk channels regulate membrane potential to mediate calcium-dependent secretion of Dpp (Fig. 6). Because we do not see a reduction in the expression or distribution of Dpp, we hypothesize that changes in the temporal dynamics of Dpp presentation account for reduced p-Mad, altered transcriptional output and disrupted morphogenesis.

Our data show that ion channels modulate cell signaling and cell fate specification in wing development in Drosophila, adding to a body of work demonstrating a role for bioelectric signaling in embryonic processes in vertebrates and invertebrates (Bates, 2015; Levin, 2012, 2014a). Indeed, mutations that disrupt Kir2.1 or the CaV1.2 calcium channel cause similar developmental phenotypes, suggesting that at least some developmental processes in vertebrates require ion channel function and could share a bioelectric mechanism (Plaster et al., 2001; Splawski et al., 2004; Swapna and Borodinsky, 2012; Tristani-Firouzi et al., 2002). Furthermore, Michael Levin's group demonstrated that disruption of bioelectric signaling can induce ectopic eyes and limbs, reverse the left-right or A/P axes, disrupt patterning of Xenopus craniofacial development, and augment regenerative properties, showing that bioelectric signaling is an important driver of morphogenesis across species (Adams et al., 2016; Levin, 2012, 2014a,b; Pai et al., 2012; Tseng et al., 2010).

Here we used Dpp-GFP as a proxy to probe the distribution of extracellular Dpp and its pattern of release, similar to studies by others (Teleman and Cohen, 2000). Endogenous Dpp forms a gradient (Raftery and Umulis, 2012), as does Dpp-GFP (Fig. 3) (Teleman and Cohen, 2000). Contrary to what is expected for reduced Dpp signaling, inhibition of Irk channels alters the gradient by expanding its width and increasing the intensity of Dpp-GFP fluorescence. At first glance our data are counterintuitive in showing that Irk2DN increases the peak intensity of Dpp-GFP (Fig. 3) and increases the baseline noise of Dpp-GFP fluorescence, but decreases p-Mad and Dpp signaling. This contradiction led us to hypothesize that Irk channels affect the temporal dynamics of Dpp exposure rather than the spatial distribution of Dpp. Using live imaging of Dpp-GFP fluorescence, we find that Dpp-GFP release occurs in bursts. Inhibition of Irk channels alters the dynamics of Dpp release, and fewer release events above baseline noise are observed (Fig. 4). However, there is more variation in the baseline fluorescence when Irk channels are inhibited, which could reflect release events that occur more constantly but with less Dpp-GFP released per event. Thus, inhibition of Irk channels would increase the frequency of Dpp release, changing Dpp release from discrete pulses to an essentially constant secretion of Dpp. Inhibition of Kir channels depolarizes cells, increases intracellular calcium and increases the secretion from excitable cells (Bondy et al., 1987; Braun et al., 2008; Cilz et al., 2014; Misler et al., 1992; Riz et al., 2015; Rorsman and Braun, 2013), and we hypothesize that a similar mechanism could be at play in the Drosophila wing disc.

We propose that alteration of Dpp dynamics reduces developmental signaling. This is consistent with the idea that temporal release of Dpp is important for developmental signaling. Siggia and colleagues used embryonic stem cells and a microfluidics in vitro system to show that the temporal pattern of exposure to a Dpp-like ligand (TGFβ) drastically affects the transcriptional readout of responding cells (Sorre et al., 2014). Compared with constant exposure to ligand, pulsed exposure to the TGFβ ligand doubles the transcriptional response and more effectively influences cell fate (Sorre et al., 2014). Here, by imaging cells in situ, we find that Dpp-GFP release also occurs in a pulsatile manner and that loss of Irk channel function alters the temporal pattern of Dpp presentation. Such a change in the temporal pattern of Dpp release would have four major consequences. First, constant secretion of Dpp would increase the amount of extracellularly localized Dpp, as we found (Fig. 3). Second, an essentially constant release of Dpp would prevent adequate refilling of vesicles with Dpp, resulting in small-amplitude Dpp-GFP release events that might escape detection and appear as a noisy baseline, consistent with our data (Fig. 5). Third, the temporal pattern of vesicular release of Dpp would change from bursts to constant, with potential effects on signaling efficacy. Fourth, reducing the function of Irk channels (by deficiency, RNAi, P-element insertion or Irk2DN expression) causes phenotypes that are typical for reduced Dpp signaling: thickened veins, crossvein defects, and reduction in wing size (Dahal et al., 2012).

Our studies reveal that Irk channel inhibition in wing discs also alters intracellular calcium activity. Whether and how these effects conspire toward a common mechanism merits further study. One possibility is that intracellular Ca²⁺ dynamics regulates Dpp release. The calcium recordings and the live imaging of Dpp-GFP release dynamics were 2 min in duration, imaging four times per second, so that we could note similarities in frequency. However, each of these measurements occurs at a different scale. The Dpp-GFP release is measured at the size of the vesicle (like a synapse), whereas the GCaMP6s signals are measured at the scale of cells. Therefore, the frequency cannot be directly compared between these experiments. That said, the average frequency of Dpp-GFP release events in control wing discs is 1.7 events per 2 min and the average number of GCaMP6s events in a cell per 2 min is close to that, at 2.3. In Irk2DN-expressing wing discs, release events have an average frequency of 1.4 events per 2 min and 1.9 GCaMP6s events per 2 min (Figs 4 and 5). Furthermore, an increase in extracellular K⁺ concentration induces Dpp release (Fig. 5). Together, these data support the hypothesis that innate changes in intracellular calcium regulate secretion of Dpp in the context of the wing disc (Fig. 6).

Membrane potential plays a key role in regulating BMP secretion in other systems. At the Drosophila neuromuscular junction, electrical activity regulates the release of a different BMP ligand, Glass bottom boat (Gbb), from postsynaptic muscle fibers to stimulate growth and development of the presynaptic boutons (Baines, 2004; Berke et al., 2013; James et al., 2014; Xiao et al., 2013). A Gbb receptor, Wit, and its downstream effector Mad are both required in the presynaptic neuron for retrograde BMP signaling to strengthen synapses (Berke et al., 2013). Interestingly, the induction of boutons by Gbb requires correct localization of functional ion-conducting glutamate receptors to the synapse (Sulkowski et al., 2014). Correctly localized but non-functional glutamate receptors do not support the effects of Gbb on boutons (Sulkowski et al., 2014). In another example, interactions between electrical activity and BMP signaling affect the balance between excitatory and inhibitory spinal neuron differentiation in the frog (Swapna and Borodinsky, 2012). Our data present the possibility that Irk channels modulate membrane potential and intracellular calcium levels to regulate BMP release from wing disc cells, as is the case in other cell types.

Several ion channels, including calcium, sodium, potassium and chloride channels, are expressed in the third instar larval discs, some of which have been shown to be specifically expressed in the wing disc (Contrino et al., 2012; Doherty et al., 1997; Klebes et al., 2002). The identification of the repertoire of ion channels that affect morphogenesis of the wing disc has the potential to reveal additional channels required for the regulation of Dpp signaling.

The Bloomington Drosophila Stock Center (BDSC) supplied many fly lines, including: UAS-Hsap-KCNJ2-GFP.EGFP7 (BDSC #6595); UAS-GFP w¹¹¹⁸; P[w⁺:UAS-GFP.nls]8 (BDSC #4776) and w[1118]; PBac{y[+mDint2] w[+mC]=2xUAS-IVS-GCaMP6s}VK (BDSC #42749). From the same source, we also obtained ry506P(PZ)hhP30, cn1P(PZ)dpp10638/CyO; ry506, tkv(y1w67c23); P(lacW)tkvk16713/(CyO) and w*; wgSp-1/CyO; P{GAL4-dpp.blk1}40C.6/TM6B, Tb1. w*,P{UAS-tkv.Q253D.Nb}3/TM3,Sb1Ser1 (Burke and Basler, 1996) was a gift from Dr R. E. Ward (University of Kansas, Lawrence, USA). hs flp; Actin>Y+>Gal4 was provided by Drs L. and Y.-N. Jan (University of California San Francisco, USA). Dr E. Ferguson (University of Chicago, USA) supplied UAS-dpp-GFP/TM3,Sb.

All flies were kept in an incubator at 25°C unless otherwise noted. Roughly equal numbers of male and female adult flies were assessed for wing defects in the rescue experiments represented in Fig. 1. All wing discs were collected from third instar larvae of ages 144-148 h after egg lay (AEL). To ensure that wing discs of both genotypes were healthy, we used terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling (TUNEL) Dead End Fluorometric TUNEL system (Promega, REFG3250) to assess apoptosis in dpp>GFP and dpp>Irk2DN; GFP wing discs. We found no apoptosis for either of these conditions (Fig. S6). In addition, to assess the health of the cells, we determined that the GFP fluorescence of dpp>Irk2DN; GFP was not obviously decreased compared with dpp>GFP wing discs, as it would be if the cytoplasm had become acidic as in a dying cell. Larvae sex was not assessed before wing disc dissection. Approximately equal numbers of male and female flies of these genotypes eclose. Therefore, we believe that both male and female wing discs are represented in all of our data.

Kcnj10-flag was amplified from a plasmid containing Flag-tagged rat Kcnj10 (gift from Dr M. Levin, Tufts University, Medford, USA) using primers: forward 5′-GCGCGGAATTCAACATGACATCAGTTGC-3′ and reverse 5′-CCGCTCGAGTCATTCACTTGTCATCG-3′. The product was cloned into a pUAST vector (Lily Jan lab, University of California, San Francisco, CA, USA) between EcoRI and XhoI sites. The construct was submitted to BestGene for generation of transgenic fly lines. For channel rescue experiments we used line eight, in which UAS-Kcnj10-flag was integrated into the third chromosome. We confirmed expression of all UAS constructs in transgenic flies with RT-PCR for Kcnj10, KCNJ2 and Irk2DN (Fig. S1).

Eggs were collected during a 4 h period. dpp>GCaMP6s (BL42749) and dpp>Irk2DN; GCaMP6s wing discs were dissected 144-148 h AEL in Shields and Sang M3 Insect Medium (Sigma-Aldrich, S3652) and transferred to Sylgard (Dow Corning) wells on glass slides containing M3 medium. The peripodial membrane of the wing discs faced the coverslip. GCaMP6s signals were acquired within 15 min of preparation at 4 Hz for 5 min, using a 488 nm laser and a Marianas inverted spinning disc microscope (Intelligent Imaging Innovations). Custom-written MATLAB (MathWorks) probes, kindly provided by Dr William J. Betz (Department of Physiology, University of Colorado Denver Anschutz Medical Campus, CO, USA), were used for analysis of calcium activity and can be obtained upon request to the corresponding author. Cells with the highest variance were identified and selected as regions of interest. The fluorescence profile for the region of interest was measured and plotted over 5 min. Calcium spikes were quantified for each disc (n=8 discs for each genotype). Samples were excluded from analysis if the wing disc moved during imaging or excessive photobleaching prevented assessment of the signal.

hs flp; Actin>Y+>Gal4, UAS-GFP/CyO virgin female flies were crossed with UAS-Irk2DN/CyO-GFP males. Clones were generated by heating the larvae for 1 h in a 37°C waterbath 1 day AEL. Third instar larvae were dissected and stained for p-Mad as described below. Fiji/ImageJ software (http://fiji.sc/Fiji) was used to measure fluorescence intensity in a rectangular region of interest inside a clone as compared with a neighboring region of the same size parallel to the A/P boundary. A two-tailed t-test was used to compare the fluorescence intensities within and outside of clones.

To control the age of the larvae, we restricted egg collection to a 4 h period. Discs were dissected 144-148 h AEL. Discs were isolated from larvae in ice-cold PBS. Isolated discs were fixed in 4% paraformaldehyde in PBS for 30 min, washed with PBS and then permeabilized using 0.3% Triton X-100 followed by treatment with 0.1% saponin. The samples were blocked in blocking buffer (3% BSA, 5% goat serum, 0.1% saponin in 1× PBS) for 1 h at room temperature. Primary antibody diluted in blocking buffer was added to the samples and incubated at 4°C overnight. Samples were washed with 0.1% saponin in PBS and incubated in secondary antibody for 2 h at room temperature and washed again with 0.1% saponin. Finally, samples were mounted in Vectashield mounting medium (Vector Labs). Primary antibodies included mouse anti-β-galactosidase [1:50; Developmental Studies Hybridoma Bank (DSHB)], rabbit anti-p-Mad (1:20; Cell Signaling, 9516S, lot 60) and rabbit polyclonal anti-DYKDDDDK-tag pAb (1:1000, Genscript, A00170). Secondary antibodies included Alexa Fluor 488 goat anti-mouse IgG (1:700; Invitrogen, A11001), goat anti-mouse IgG Cy5 (1:200; Millipore, AP124S) and goat anti-rabbit IgG Cy3 (1:700; Invitrogen, A10520, lot 1387825). Imaginal discs from third instar MS1096>KCNJ2-GFP control and MS1096>Irk2DN; KCNJ2-GFP larval discs were stained with mouse monoclonal GFP antibody (1:200; Santa Cruz Biotechnology, SC-101525) and rat DCAD2 antibody (1:100; DSHB). MS1096>Kcnj10-flag and MS1096>Irk2DN; Kcnj10-flag discs were stained with mouse anti-Flag (1:500; Sigma-Aldrich, F1804-50UG, lot SLBK1345V) and rat anti-DCAD2 (1:100; DSHB). The wing discs were co-stained with secondary antibodies conjugated with Alexa Fluor 488 and Cy5.

The rabbit anti-Dpp prodomain antibody (Akiyama and Gibson, 2015) (gift from Matthew Gibson, Stowers Institute, Kansas City, USA) was diluted 1:100 in Can Get Signal B (Toyobo) containing 0.1% Triton X-100, 10% normal goat serum and anti-rabbit Cy3 (1:500; Life Technologies). The discs were mounted in PermaFluor mountant (ThermoScientific). Confocal z-stack images of Dpp-stained wing discs were taken with a Zeiss LSM 780 confocal microscope. Control and experimental samples were imaged on the same day with the same settings. The fluorescence intensity was measured from posterior to anterior in four lines across the A/P boundary, with two above and two below the D/V boundary using Fiji/ImageJ software. To average fluorescence intensity profiles of all samples, the central Dpp-stained region was aligned at the peak intensity. Data were normalized to the minimum baseline intensity.

For staining extracellular Dpp-GFP, third instar wing discs were collected from larvae at 130-134 h AEL in Schneider 2 culture medium (Life Technologies) and incubated in mouse monoclonal anti-GFP (1:50) for 1 h on ice. Discs were washed with PBS and fixed in 4% paraformaldehyde. Subsequent staining procedures followed previously described protocols (Strigini and Cohen, 2000). Alexa Fluor 488 goat anti-mouse IgG (1:700; Invitrogen) was used as the secondary antibody.

Control and experimental wing discs were imaged on the same day with the same settings. The maximum intensity projection was used for fluorescence intensity analysis using Fiji/ImageJ software. Immunofluorescence of wing discs was measured in four lines across the disc immediately dorsal of the D/V boundary as depicted in Fig. 2A. Fluorescence was measured in four lines for each disc from posterior to anterior across the wing disc. To normalize to the background, we used the anchor point method, setting the anchor point as the minimum of the posterior trough of fluorescence intensity (Brooks et al., 2012).

Wing discs from third instar dpp>dpp-GFP and dpp>Irk2DN; dpp-GFP larvae (144-148 h AEL) were isolated and placed in PBS on a Premium cover glass (Fisher Scientific) with the peripodial side facing down. Maximum intensity projection images were acquired from confocal z-stacks (Zeiss LSM 780 confocal microscope) before and after addition of ammonium chloride to a final concentration of 20 mM. The same settings were used to acquire confocal images before and 1 min after addition of ammonium chloride.

Wing discs were dissected and transferred to Sylgard wells containing HL3.1 saline solution (70 mM NaCl, 5 mM KCl, 1.5 mM CaCl₂, 4 mM MgCl₂, 10 mM NaHCO₃, 5 mM trehalose, 115 mM sucrose and 5 mM HEPES, pH 7.4) (Jan and Jan, 1976; Stewart et al., 1994) with the peripodial side facing away from the slide. Thus, upon applying a coverslip, the disc can be imaged from the apical side. Images were taken with a Zeiss LSM 780 confocal microscope every 250 ms over 120 s. Control and experimental samples were imaged using the same confocal settings. For analysis of Dpp-GFP fluorescence, images were loaded into MATLAB as uncompressed AVI files and analyzed using custom-written programs (kindly provided by Dr William J. Betz). Two regions of 50×50 pixels were selected for analysis. The first was just anterior to the Dpp-producing A/P boundary and ventral of the D/V boundary. The second was just anterior to the ventralmost Dpp-producing cells. Pixels with the highest variance in fluorescence throughout the file were identified and selected as a region of interest with a radius of three pixels. The fluorescence profile of the brightest pixel within the region of interest was plotted over time. Release events were identified as a rise in fluorescence significantly above baseline. Imaging files were excluded from analysis if the wing disc moved during imaging.

Wing imaginal discs from 144-148 h AEL dpp>dpp-SEP third instar larvae were dissected and placed in a well containing HL3.1 saline solution on a 24×40 mm coverslip with the peripodial side facing down. Confocal images were taken every 250 ms for 500 cycles. While the images were being taken, HL3.1 saline containing 1 M potassium chloride was added with a long slender gel-loading pipette tip so that the final concentration was 100 mM potassium (Shui et al., 1998). We noted the cycle at which the high potassium solution was added. Control experiments were performed with the addition of normal HL3.1 saline during the imaging. Imaging files were excluded from analysis if the wing disc moved during imaging.

Wing discs were isolated from third instar larvae and total RNA extracted using the SV Total RNA Isolation Kit (Promega). Single-stranded cDNA was synthesized using the Superscript III First Strand Synthesis System (Invitrogen). We used a LightCycler 480 with SYBR Green I Master Mix (Roche Diagnostics) and the following primers (5′-3′) to amplify specific transcripts: Rp49 (RpL32), AAGAAGCGCACCAAGCACTTCATC and TCTGTTGTCGATACCCTTGGGCTT; KCNJ2, GATGTGTCACGGATGAATGC and ACTCGCCACATCAAACACAG; Kcnj10, CAAGGTCTATTACAGCCAGAC and TCCATAGATCCTTGAGGTAGAG; Ork, TCCTGCTGCTCATCTTCTA and AGCTCCTCCAGCAGATATT; Irk2DN, ACACTATTGCCGCTGCC and CAAATTACGGCGTGTTTGGAGAAC.

We thank Dr Lee Niswander for insightful conversations and suggestions; Megan Josey for copy editing; Dr Angeles Ribera and Dr Rosa Moreno for suggestions, discussion, attempts to measure resting membrane potential and explore electrical activity in the wing disc; Dr William J. Betz for writing and assisting in the use of custom MATLAB programs; the University of Colorado Anschutz Medical Campus Advanced Light Microscopy Core and Dr Radu Moldovan for training and use of the 3i Marianis spinning disc confocal; and Matthew Gibson (Stowers Institute) for the Dpp prodomain antibody. The mouse anti-β-galactosidase antibody was developed by Joshua Sanes, and the rat DCAD2 antibody was developed by T. Uemura, and both were obtained from the Developmental Studies Hybridoma Bank created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, USA. We thank the Bloomington Drosophila Stock Center for many fly stocks and Dr Michael Levin for the plasmid with cDNA encoding Flag-tagged Kir4.1. We thank the Nancy and Michael Burgermeister Family who donated the funds for the purchase of the Zeiss LSM 780 microscope.